MCP Setup Guide - Claude Desktop Integration
Connect Kagura to Claude Desktop in 2 minutes
This guide shows how to integrate Kagura's universal memory with Claude Desktop using the Model Context Protocol (MCP).
📋 Prerequisites
- Kagura AI v4.0+ installed
- Claude Desktop (supports MCP)
⚡ Automatic Setup (Recommended)
Kagura can automatically configure Claude Desktop for you:
Output:
✅ Successfully installed!
Configuration:
Server name: kagura-memory
Command: kagura mcp serve
Config file: ~/.config/claude/claude_desktop_config.json
Next steps:
1. Restart Claude Desktop
2. Start a new conversation
3. Try: 'Remember that I prefer Python'
That's it! Kagura is now connected to Claude Desktop.
🔧 Manual Setup (Alternative)
If automatic setup doesn't work, you can manually edit the config file.
Step 1: Locate Claude Desktop Config
macOS/Linux:
Windows:
Step 2: Edit Configuration
Add Kagura to the mcpServers section:
Full example:
{
"mcpServers": {
"kagura-memory": {
"command": "kagura",
"args": ["mcp", "serve"],
"env": {}
},
"other-server": {
"command": "other-command",
"args": ["serve"]
}
}
}
Step 3: Restart Claude Desktop
Close and reopen Claude Desktop to apply changes.
✅ Verify Integration
Method 1: Ask Claude
Start a new conversation in Claude Desktop and try:
You: "Remember that I prefer Python over JavaScript for backend projects"
Claude will use the memory_store tool to save this.
You: "What programming languages do I prefer?"
Claude will use memory_recall or memory_search to retrieve the information.
Method 2: Check Diagnostics
Look for:
🧠 Available Memory Tools
Once integrated, Claude has access to these memory tools:
Core Tools
| Tool | Purpose | Example |
|---|---|---|
| memory_store | Save information | "Remember X" |
| memory_recall | Get by key | "What did I say about Y?" |
| memory_search | Semantic search | "Find memories about Z" |
| memory_list | List all memories | "What do you remember about me?" |
| memory_feedback | Mark useful/outdated | Automatic |
| memory_delete | Forget information | "Forget about X" |
Memory Scopes
- working: Temporary, session-only (default)
- persistent: Saved to disk, survives restart
Example Interactions
Store persistent memory:
"Remember that my favorite Python library is FastAPI. This is important and should be persistent."
Search memories:
"What do you know about my coding preferences?"
Feedback (automatic):
Claude automatically marks memories as "useful" when they help answer your questions.
Delete:
"Forget about my old JavaScript preference"
🔍 Troubleshooting
Claude Desktop doesn't see Kagura tools
Check 1: Verify installation
Check 2: Restart Claude Desktop - Quit Claude Desktop completely - Reopen it - Start a new conversation
Check 3: Check logs
"kagura command not found"
Solution: Use full path in config
{
"mcpServers": {
"kagura-memory": {
"command": "/full/path/to/kagura",
"args": ["mcp", "serve"]
}
}
}
Find full path:
Memory not persisting across conversations
Cause: Using scope="working" (default)
Solution: Explicitly use scope="persistent"
Or tell Claude:
"Remember this permanently: I prefer Python"
🚫 Uninstall
To remove Kagura from Claude Desktop:
This removes the configuration but does not delete your stored memories.
🔗 Related
- Getting Started - Installation guide
- API Reference - REST API docs
- Architecture - System design
Version: 4.0.0a Last updated: 2025-10-26